Hypertext systems have been become popular in recent years although the ideas behind them were proposed nearly fifty years ago. Hypertext links, which are connections between information items, provide the possibility for non-sequential reading and writing, and enable related pieces of information to be connected together no matter where they are stored in a system. Traditionally, a hypertext link is created by specifying its two ends manually. Links created in this way can be excellent but suffer from limitations in that they require manual work. In a system where information is frequently changed or a huge amount of information is stored, these links may become inadequate. Methods that can create links automatically are needed. In this thesis, two kinds of computer created link, retrieval-links and doc-links, are proposed and investigated. Discussion are focused on how to control link creation and how to design proper full text retrieval procedures for link creation. Dynamically dividing documents into sections is an example of enabling the user to control link creation, and the use of break words and source stems are products of exploring effective retrieval methods. With efforts to ensure effectiveness and efficiency in link creation, the proposed methods can provide extra support for users and require virtually no extra human effort or user interface skills. In this thesis, key issues in link creation are identified, problems and solutions are discussed, and systematic experiments and their results are presented and analysed. Based on these results, the proposed methods are implemented and integrated with a hypermedia system, Microcosm. Practice has proved that these methods are effective and useful. Possible improvements and issues for developing better information systems are discussed as well.